A Safeguarded Projected-Gradient Framework for Complementarity-Constrained Least-Squares Problems

Generalized absolute value equations (GAVEs) and linear complementarity problems (LCPs) arise in many equilibrium and optimization models and can both be formulated as least-squares problems over the complementarity set.

Building on this shared formulation, we propose a safeguarded projected-gradient framework for both classes with problem-dependent algebraic refinements.

We analyze its convergence properties and, under stronger conditions, establish finite recovery of the unique solution.

Numerical experiments on GAVE and LCP benchmarks illustrate the practical behavior of the framework, including its high final accuracy in the tested settings and its problem-dependent computational tradeoffs.